E-LITMUS

Executive Summary

e-Litmus is a universal health monitoring system for EV charging infrastructure, providing charger-agnostic predictive maintenance across all major brands (ABB, Tritium, Kempower, Alpitronic, etc.).

The platform combines 7-sensor hardware (RF arcing detection, thermal imaging, barometric pressure, current/voltage, CAN/Modbus logging, vibration, environmental) with cloud ML analytics for real-time anomaly detection (4-hour response window) and IGBT degradation trending (weeks-to-months advance warning) R39R40, enabling CPOs to shift from reactive firefighting to planned maintenance. [Karanam & Tal, EVS36, 2023; Nazar et al., PHM Europe 2020]

Value proposition: 25% downtime reduction R12, 30-35% maintenance cost savings R2, and uptime-driven revenue growth. Pilot-ready for Q2 2025 with first 3 CPOs.

The Problem

EV Charger Downtime Crisis

Charge Point Operators (CPOs) face a critical challenge: unpredictable charger failures that reduce uptime and increase operational costs. Every hour of downtime means lost revenue, frustrated customers, and emergency repair expenses.

The global predictive maintenance market is projected to reach $47.8B by 2029 R1, driven by the urgent need to minimize unplanned downtime across industrial assets. For EV charging infrastructure, this challenge is particularly acute: charger availability directly impacts revenue, as every minute of downtime translates to lost charging sessions.

The Financial Impact of Downtime

According to Siemens' True Cost of Downtime 2024 report R2, unplanned industrial asset failures cost enterprises an average of 11% annual revenue. For CPOs, this translates to millions in lost opportunity.

CPO Operational Challenges

Modern Charge Point Operators manage complex, multi-vendor fleets across distributed locations. Key challenges include:

• Mixed Fleet Complexity: Most CPOs deploy 3-5 different charger brands (ABB, Tritium, Kempower, Alpitronic, Delta) with incompatible monitoring systems
• Blind Spots: OEM dashboards only show transactional data (kWh delivered, session status), not asset health indicators
• Reactive Fire-Fighting: Maintenance teams respond to customer complaints rather than preventing failures
• False Alarms: Generic alerts flood operations centers without actionable diagnostics
• No Root Cause Analysis: Technicians replace entire modules (€14K-125K each) without understanding what failed

Why Traditional Monitoring Fails

• Reactive Approach: OEM dashboards show failures after they happen, not before
• Single-Vendor Lock-In: ABB Ability, Kempower Cloud only monitor their own chargers
• Transactional Focus: CSMS platforms (Hubject, ChargePoint) manage payments, not asset health
• No Predictive Intelligence: Basic alerts without failure prediction or root cause analysis

The Solution

Charger-Agnostic Predictive Maintenance

e-Litmus monitors EV charger health through non-invasive sensors and 2-level analytics — no charger manufacturer lock-in.

Inspired by successful predictive maintenance implementations in industrial sectors (see Tractian, Artesis), e-Litmus brings AI-driven asset health monitoring to EV charging infrastructure. Unlike SKF's motor monitoring systems or Sensemore's vibration analytics, e-Litmus is purpose-built for the unique failure modes of DC fast chargers.

7-Sensor Suite with Full Specifications

Non-Invasive Installation: All sensors mount externally on charger cabinet. No modification to charger internals required. Installation time: 2-4 hours/charger with standard tools. Compatible with all DC fast charger enclosures (IP54/IP55 rated).

Edge Computing Module

e-Litmus uses industrial-grade edge gateways powered by NXP i.MX RT1170 or Infineon AURIX platforms for on-device machine learning inference.

GSM Connectivity

The edge gateway transmits aggregated health metrics via 4G LTE every 10 minutes (configurable: 5-60 min). Data payload: ~2 KB/transmission = ~288 KB/day/charger. Annual bandwidth: ~105 MB/charger (minimal data costs).

• Primary: 4G LTE Cat-M1 (low-power cellular for IoT)
• Fallback: NB-IoT (ultra-low power, wider coverage)
• Backup: Ethernet (if available at site)

Network carriers: Compatible with global LTE networks (Vodafone, T-Mobile, AT&T, etc.). eSIM support for multi-region deployments.

Technology Highlight: 2-Level Analytics

Level 1: Edge Analytics (Per-Charger):

• Real-Time Anomaly Detection: Isolation Forest algorithm runs on-device, detects deviations from learned baselines
• Latency: <10ms inference time, immediate local alerts
• Training: Self-learning during first 30 days of "normal" operation, then continuous model updates
• Local Actions: Critical alerts trigger immediate notifications (SMS, email) without cloud dependency
• Privacy: Raw sensor data never leaves device; only aggregated health scores transmitted

Level 2: Backend Analytics (Fleet-Wide):

• Remaining Useful Life (RUL) Prediction: LSTM neural networks trained on 10,000+ charger-hours of telemetry
• Failure Classification: Supervised ML models categorize 15 failure modes (power module, cooling, grid, cable, connector, etc.)
• Cross-Fleet Benchmarking: Compare charger health across sites, identify underperforming assets
• Predictive Horizon: Real-time anomaly detection (4-hour response); IGBT degradation trending: weeks-to-months advance warning R39R40. Note: EV charger PdM is an emerging field — e-Litmus pioneers data collection for future RUL prediction
• Integration: RESTful API for CSMS platforms (Hubject, ChargePoint, custom dashboards)

Inspired by deep learning approaches for predictive maintenance (Nature Scientific Reports, 2020) R39 and machine learning for industrial asset monitoring (Nature, 2023) R40.

Architecture

VPS-Based Cloud Platform (4-Server Architecture)

e-Litmus deploys on a lean, cost-effective 4-VPS cluster architecture (pilot scale, 1000 chargers), with horizontal scalability to 6-8 VPS for production (10K+ chargers). Inspired by modern cloud-native patterns (see AWS IoT Core architecture), but deployed on VPS infrastructure for cost control and data sovereignty.

Total Infrastructure Cost: ~€140-185/month for 4 VPS servers (Hetzner, OVH, or similar providers). Scales to 1000 chargers without additional hardware.

Data Flow Architecture

Multi-Tenancy Architecture (Keycloak)

e-Litmus supports multi-tenant SaaS deployment using Keycloak realms for data isolation. Each CPO (Tenant-1: Ionity, Tenant-2: Fastned, Tenant-3: Milence) operates in a separate security realm with dedicated JWT signing keys and database partitions.

Tenant Isolation Mechanisms:

• Authentication: Each tenant has dedicated Keycloak realm, client_id, client_secret → JWT tokens include tenant_id claim
• Database: PostgreSQL schemas enforce row-level security (RLS) policies → Queries filtered by tenant_id
• MQTT: Topic hierarchy enforces tenant isolation (tenant_id/box_id/sensors) → Processing service validates tenant_id before processing
• API Gateway: Nginx validates JWT before routing → Ingestion service rejects cross-tenant requests (returns 403 Forbidden)
• Audit Logging: All cross-tenant access attempts logged to security audit log → Compliance with GDPR, ISO 27001

Firmware Configuration (Edge Gateway)

Each telemetry box is pre-configured with tenant-specific credentials during manufacturing/provisioning:

KEYCLOAK_CONFIG = {
    "realm": "tenant-1",
    "client_id": "tenant-1-box",
    "client_secret": "[client-secret]",
    "auth_url": "https://api.platform.com/auth/realms/tenant-1/protocol/openid-connect/token",
    "api_endpoint": "https://api.platform.com/api/v1/ingest"
}

# Device sends telemetry every 10 minutes
TELEMETRY_INTERVAL_SEC = 600  # 10 minutes

# Edge ML model (TFLite Micro, runs on NXP i.MX RT1170)
EDGE_ML_MODEL = "isolation_forest_v2.tflite"
ANOMALY_THRESHOLD = 0.85  # 85% confidence for alert trigger
                

Scalability & Performance

Infrastructure Monitoring: Prometheus scrapes metrics from all services (Nginx, MQTT Broker, PostgreSQL, Ingestion/Processing services). Grafana dashboards visualize system health, ingestion rates, processing delays, database performance. Alerts via email/SMS for critical failures.

Security & Compliance

• TLS 1.3: All API traffic encrypted in transit (HTTPS, WSS for WebSocket)
• At-Rest Encryption: PostgreSQL transparent data encryption (TDE), encrypted disk volumes on VPS-3
• OAuth2/JWT: Industry-standard authentication via Keycloak (supports SAML, OIDC for enterprise SSO)
• Role-Based Access Control (RBAC): CPO admin, technician, viewer roles with granular permissions
• Audit Logging: Tamper-proof audit trail for all API access, data modifications, cross-tenant attempts
• GDPR Compliance: Data residency controls (EU-only VPS deployment option), right-to-erasure APIs, consent management
• Penetration Testing: Annual third-party security audits, vulnerability scanning (OWASP Top 10)

Technology Stack Summary

Business Value & ROI

Quantified Benefits

• 25% downtime reduction: From reactive firefighting to planned maintenance windows R12
• 30-35% maintenance cost savings: Predictive parts replacement vs. emergency repairs R2
• Revenue protection: Each hour of charger downtime = €50-150 lost revenue (fast charging rates)
• Customer satisfaction: Proactive maintenance builds trust, reduces complaints
• Regulatory compliance: Meet uptime SLAs for government subsidies and fleet contracts

Real Diagnostic Case Study

Incident: "Charger Stopped at 60%" — Avoiding €115K Misdiagnosis

Location: Highway charging hub, ABB Terra 184 DC fast charger (180 kW rated)

Date: July 2024 (e-Litmus pilot deployment)

Timeline of Events

Sensor Data Analysis

Graph 1: Barometric Pressure (09:00-09:30):

• Baseline (09:00-09:10): 997.0 hPa (normal atmospheric pressure + cabinet positive pressure from fans)
• Anomaly (09:12): Sudden drop to 994.0 hPa (-3.0 hPa = significant airflow restriction)
• Indicator: Air filter clog → reduced airflow → pressure drop inside cabinet
• e-Litmus Alert: 🔴 Anomaly Detection triggered (Isolation Forest model, 92% confidence)

Graph 2: Internal Cabinet Temperature (09:00-09:30):

• Baseline (09:00-09:10): 25°C (ambient 20°C + normal heat from power modules)
• Spike (09:12-09:20): 25°C → 60°C (abnormal thermal buildup)
• Correlation: Restricted airflow → insufficient cooling → thermal runaway
• Result: Charger enters thermal protection mode (automatic power de-rating to prevent damage)

Graph 3: Charging Current (09:00-09:30):

• Start (09:02): 0A → 180A (150 kW @ 830V DC)
• De-rating (09:16): 180A → 120A (thermal protection activates)
• Further de-rating (09:18): 120A → 60A (critical thermal threshold)
• Session end (09:20): 60A → 0A (charger shuts down to prevent damage)

Graph 4: Acoustic Sensor (Fan Speed Analysis):

• Fan RPM: Remained constant at 2800 RPM (normal speed, fans operational)
• Conclusion: Fans working correctly, but airflow blocked externally (filter clog, not fan failure)

Competing Diagnoses

Resolution & Impact

Action Taken (based on e-Litmus diagnosis):

1. Technician dispatched within 30 minutes (local maintenance crew)
2. Air filter inspection revealed 80% clog (dust, pollen accumulation over 6 months)
3. Filter cleaned in 10 minutes (no parts replacement needed)
4. Charger tested: internal temperature returned to 25°C, full 180 kW power output restored
5. Customer notified: "Technical issue resolved, free charging voucher as apology"

Financial Impact:

Additional Benefits:

• Warranty Protection: Avoided invalid warranty claim (ABB would reject €115K claim after discovering filter clog was preventable maintenance issue)
• Customer Retention: Proactive communication turned complaint into loyalty (free charging voucher cost: €45)
• Knowledge Base: e-Litmus ML model learned new failure pattern → improved future diagnostics for entire fleet
• Preventive Maintenance: CPO scheduled quarterly filter inspections across all 1000 chargers → prevented 50+ similar incidents/year

Competitive Positioning

Unique Value Proposition

"The only charger-agnostic predictive maintenance platform for CPOs, reducing OPEX 7-10% and increasing revenue 10% through uptime optimization. 2-level analytics: edge (per-charger) + backend (fleet-wide benchmarking)."

vs. CSMS Platforms (Hubject, ChargePoint)

• Focus: Predictive asset health vs. transactional charging management
• Value: Uptime optimization vs. payment processing
• Integration: API for CSMS platforms (partner opportunity)

vs. Charger OEM Dashboards (ABB Ability, Kempower Cloud)

• Multi-Vendor: Charger-agnostic vs. single-manufacturer lock-in
• Analytics: Predictive ML vs. reactive alerts
• Coverage: Mixed fleets (CPOs use 3-5 OEMs) vs. single-brand monitoring

vs. Generic IoT Platforms

• Domain Expertise: EV charging workflows (OCPP integration, CPO KPIs)
• Turnkey: Pre-trained failure mode libraries vs. custom ML development
• Pricing: SaaS (€195K Y1 → €23K Y2+) vs. platform fees + development costs

Business Model

Pure SaaS Subscription

e-Litmus operates on a SaaS subscription model with transparent, predictable pricing R8. Year 1 includes hardware deployment, integration, and 12-month SaaS access. Year 2+ is pure software subscription with minimal operational costs.

Flexible Deployment Options:

• Full Fleet (1000+ chargers): €195K Year 1 → €23K/year thereafter
• Pilot (50-100 chargers): €14K Year 1 → €3K/year thereafter
• Single Site (10-20 chargers): €5K Year 1 → €1K/year thereafter

Target Customers

• Primary: CPOs (Charge Point Operators) — Ionity, Fastned, Milence, Electra, Epic Charging, Charge&Go
• Secondary: CSMS providers (Charging Station Management Systems) — Hubject, ChargePoint, Gireve — white-label integration for asset health layer
• Tertiary: DC fast charger OEMs — ABB, Tritium, Kempower, Alpitronic, Delta — licensing model for built-in predictive maintenance
• Adjacent: O&M (Operations & Maintenance) service providers — companies managing charger fleets for third parties

ROI Analysis: 1000-Charger Network

Baseline Assumptions (typical CPO with 1000 DC fast chargers):

Revenue Impact: +2% Uptime Improvement

e-Litmus predictive maintenance reduces unplanned downtime from 8% to 6% (2% uptime gain) through early failure detection and scheduled repairs.

OPEX Impact: -30% Maintenance Cost Reduction

Predictive maintenance shifts repair work from emergency reactive to scheduled proactive, reducing labor costs, parts waste, and unnecessary component replacements.

Conservative OPEX Savings Estimate: €225K-495K/year (excluding emergency call-out savings and parts waste prevention).

Year 1 ROI Calculation

Payback Period: 3-4 months (assuming conservative scenario). R1R2

Year 2+ Economics

Year 2 and beyond: e-Litmus subscription cost drops to €23K/year (pure SaaS, no hardware costs). Benefits remain constant as long as fleet is maintained.

5-Year Total Net Benefit: €610K (Year 1) + €782K (Year 2) + €782K (Year 3) + €782K (Year 4) + €782K (Year 5) = €3.74M (conservative scenario).

Competitive Pricing Benchmark

Implementation Roadmap

e-Litmus follows a phased market entry strategy, starting with pilot deployments to validate ROI, then scaling through direct CPO sales, CSMS partnerships, and OEM licensing.

Phase 1: Proto (July 2024 - December 2024)

Objective: Solution prototyping, customer development, first pilot deployment

Key Partners:

• Zynetic (CPO, pilot partner)
• NXP Semiconductors (hardware platform provider)
• Startup accelerators (TechStars Energy, Plug and Play Energy)

Phase 2: MVP (2025)

Objective: Legal incorporation, seed funding, expand pilot deployments, validate commercial model

Target Customers (Pilot → Commercial):

• Tier 1 CPOs: Ionity, Fastned, Milence (1000+ charger fleets) — €195K contracts
• Tier 2 CPOs: Electra, Epic Charging, Atlante, Charge&Go (200-500 chargers) — €45K-90K contracts
• CSMS Platforms: Hubject, Gireve — white-label integration discussions

Phase 3: Traction & Capitalization (2026-2027)

Objective: Commercial deployments, sustainable subscription revenue, platform partnerships, OEM licensing

Geographic Expansion:

• Europe: Norway, Germany, France, Netherlands, UK (primary market)
• North America: US (California, Texas, Florida), Canada (Ontario, Quebec) — pilot deployments

Phase 4: Scaling (2027-2030)

Objective: Sales scaling, Series A funding, M&A exit options

• Series A Funding: €5-10M (target: 2027) → accelerate US expansion, R&D for V2 platform
• Target Customers: 50+ CPOs, 50,000+ chargers under management
• Exit Options:
  • Strategic M&A: Acquisition by CSMS platform (Hubject, ChargePoint, Driivz) or charger OEM (ABB, Tritium, Kempower)
  • IPO: If sustainable revenue trajectory (>€18M ARR by 2029)

Key Sales Channels

Competitive Moat

Why e-Litmus is defensible:

• Domain Expertise: 10,000+ charger-hours of telemetry data → pre-trained ML models (barrier for new entrants)
• Multi-Vendor Integration: Works with all charger brands (vs. OEM lock-in), CSMS-agnostic APIs
• Network Effects: More chargers → better ML models → higher diagnostic accuracy → stickier customers
• IP Protection: Patents pending on multi-sensor fusion for EV charger diagnostics (filed Q3 2024)
• First-Mover Advantage: No direct competitor offers charger-agnostic predictive maintenance SaaS today

Promwad Competencies

Promwad brings 20+ years of embedded systems and IoT platform development expertise to e-Litmus:

Relevant Experience

• Automotive Telematics: 300K+ units shipped (4G LTE + GPS + CANBUS integration), proven at scale
• Industrial IoT: Vibration monitoring systems, edge ML inference, predictive maintenance platforms
• EV Ecosystem: Experience with automotive communication protocols (CAN, Modbus, OCPP)
• Multi-Tenant SaaS: AWS IoT Core, PostgreSQL RLS, Keycloak authentication for B2B platforms
• Edge ML: TensorFlow Lite Micro, on-device anomaly detection, LSTM RUL prediction models

Delivery Capabilities

• Hardware Design: Industrial-grade IoT sensors, IP65 enclosures, RF/thermal/vibration integration
• Firmware Development: Real-time data acquisition, edge ML inference, OTA updates
• Cloud Platform: VPS architecture, PostgreSQL+TimescaleDB, MQTT broker, React dashboards
• ML Pipeline: LSTM (RUL prediction), Random Forest (classification), Isolation Forest (anomaly detection)
• Certifications: ISO 9001, ASPICE Level 2, IEC 62304 (medical devices)
• Official Partnerships: NXP, Qualcomm, Qt, Nordic Semiconductor

External Links

• Promwad: Automotive IoT Solutions — Engineering expertise in automotive electronics
• Promwad: Company Profile — Hardware/embedded systems development partner

Ready to Start?